Reproduce “Monotone deep Boltzmann machines” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zhili Feng, Ezra Winston, J. Zico Kolter
The mDBM uses constrained weights to retain within-layer connections in a Boltzmann machine while enabling reliable parallel mean-field inference. Under a monotonicity condition, its mean-field solution is unique and globally optimal, and a damped parallel procedure converges to it. On MNIST with 60% of pixels masked, it achieves 92.95% classification accuracy, versus 64.23% for an RBM tested with mean-field inference.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.